Improved Bioproduction of the Nylon 12 Monomer by Combining the Directed Evolution of P450 and Enhancing Heme Synthesis

The nylon 12 (PA12) monomer ω-aminododecanoic acid (ω-AmDDA) could be synthesized from lauric acid (DDA) through multi-enzyme cascade transformation using engineered E. coli, with the P450 catalyzing terminal hydroxylation of DDA as a rate-limiting enzyme. Its activity is jointly determined by the heme domain and the reductase domain. To obtain a P450 mutant with higher activity, directed evolution was conducted using a colorimetric high-throughput screening (HTS) system with DDA as the real substrate. After two rounds of directed evolution, a positive double-site mutant (R14R/D629G) with 90.3% higher activity was obtained. Molecular docking analysis, kinetic parameter determination and protein electrophoresis suggested the improved soluble expression of P450 resulting from the synonymous mutation near the N-terminus and the shortened distance of the electron transfer between FMN and FAD caused by D629G mutation as the major reasons for activity improvement. The significantly increased kcat and unchanged Km provided further evidence for the increase in electron transfer efficiency. Considering the important role of heme in P450, its supply was strengthened by the metabolic engineering of the heme synthesis pathway. By combining P450-directed evolution and enhancing heme synthesis, 2.02 ± 0.03 g/L of ω-AmDDA was produced from 10 mM DDA, with a yield of 93.6%.


Introduction
As an important engineering plastic, nylon 12 (PA12) has a number of advantages such as low water absorption, a high dimensional stability and a high temperature resistance and corrosion resistance and thus has wide applications in automobiles, electrical appliances, aerospace, etc. [1,2]. At present, the industrial production of PA12 mainly adopts the oxidation process, with butadiene as the raw material [3], but this process has problems such as the use of toxic and harmful raw materials, the need for a high reaction temperature, the dependence on nonrenewable petrochemical raw materials and the environmental stress caused. In contrast, the biosynthesis of the PA12 monomer from renewable resources is a green and sustainable process with mild reaction conditions, and it has thus emerged as a promising alternative.
Using methyl laurate as a raw material, the biosynthesis of 12-aminododecanoic acid methyl ester (ADAME) was realized through whole-cell catalysis with a yield of 12% (129 mg/L) [4]. The biotransformation of lauric acid (DDA) to ω-aminododecanoic acid (ω-AmDDA) was first reported in 2018 by using a mixture of two engineered strains, with a yield of 30% (93 mg/L) [5]. Recently, we constructed a cofactor self-efficient E. coli strain through the design of cofactor regeneration cycles and metabolic engineering of the chassis cell, which produced 1.04 g/L of ω-AmDDA from DDA at a yield of 96.5% [6]. In the The heme domain of P450 catalyzes the selective oxidation of inert hydrocarbon bonds through correct binding with the substrate. This reaction process relies on the coenzyme NAD(P)H and the complex electron transfer chain system. The shape and size of the substrate binding pocket, the efficiency of the electronic transfer chain system and the adequate supply of heme all contribute to the activity of P450. At present, most studies on P450 engineering focus on the heme domain [7][8][9]. For example, the G307A mutant of CYP153a from Marinobacter aquaolei enhanced the activity of the chimeric P450 enzyme (cyp153a-ncp) towards fatty acids by 2-to 20-fold. In recent years, there have also been a few reports on the engineering of the P450 reductase domain [8,10] and the interface between the heme domain and the reductase domain [8]. For example, the S120R/P165N/S453N mutant of CYP153a improved the electron transfer efficiency of redox partners to CYP153a, which increased its ω-hydroxylation activity towards oleic acid by 2.7-fold [8]. In comparison to rational or semi-rational design targeting selected residues, the directed evolution of the whole protein covering all three regions may generate mutants with higher activity.
For efficient directed evolution, an appropriate high-throughput screening (HTS) method is a necessity to facilitate the accurate selection of target mutants with desirable features from the huge random mutant library [11]. Diammonium 2,2′-azino-bis(3ethylbenzothiazoline-6-sulfonate) (ABTS) colorimetry [12] has been used to screen P450 mutants with a high ω-hydroxylation activity of fatty acids. This method was developed based on the specific oxidation activity of a galactose oxidase mutant (GOaseM3-5) towards terminal fatty acid hydroxylates, which can be coupled with the colorimetric measurement of H2O2 generated during this reaction. Therefore, this method may be suitable for the screening of P450 mutants with enhanced terminal hydroxylation activity of DDA. However, the Fe 2+/3+ and D-glucose required for P450 catalysis and NADPH regeneration in the whole-cell reaction of DDA hydroxylation may interfere with the detection results of GOaseM3-5 [13]. If the Fe 2+/3+ /Cu 2+ ratio in the solution is too high, the Cu 2+ binding of GOaseM3-5 would be hindered, leading to a loss of activity. Meanwhile, the activity of GOaseM3-5 towards D-glucose may generate false positive results. Therefore, this HTS method needs further improvement before it can be applied for P450-directed evolution in the whole-cell reaction system.
Moreover, the activity of catalytic C-H activation is mediated by the active P450 formed after the binding with heme. In the catalytic process, heme plays the key role of The heme domain of P450 catalyzes the selective oxidation of inert hydrocarbon bonds through correct binding with the substrate. This reaction process relies on the coenzyme NAD(P)H and the complex electron transfer chain system. The shape and size of the substrate binding pocket, the efficiency of the electronic transfer chain system and the adequate supply of heme all contribute to the activity of P450. At present, most studies on P450 engineering focus on the heme domain [7][8][9]. For example, the G307A mutant of CYP153a from Marinobacter aquaolei enhanced the activity of the chimeric P450 enzyme (cyp153a-ncp) towards fatty acids by 2-to 20-fold. In recent years, there have also been a few reports on the engineering of the P450 reductase domain [8,10] and the interface between the heme domain and the reductase domain [8]. For example, the S120R/P165N/S453N mutant of CYP153a improved the electron transfer efficiency of redox partners to CYP153a, which increased its ω-hydroxylation activity towards oleic acid by 2.7-fold [8]. In comparison to rational or semi-rational design targeting selected residues, the directed evolution of the whole protein covering all three regions may generate mutants with higher activity.
For efficient directed evolution, an appropriate high-throughput screening (HTS) method is a necessity to facilitate the accurate selection of target mutants with desirable features from the huge random mutant library [11]. Diammonium 2,2 -azino-bis(3ethylbenzothiazoline-6-sulfonate) (ABTS) colorimetry [12] has been used to screen P450 mutants with a high ω-hydroxylation activity of fatty acids. This method was developed based on the specific oxidation activity of a galactose oxidase mutant (GOase M3-5 ) towards terminal fatty acid hydroxylates, which can be coupled with the colorimetric measurement of H 2 O 2 generated during this reaction. Therefore, this method may be suitable for the screening of P450 mutants with enhanced terminal hydroxylation activity of DDA. However, the Fe 2+/3+ and D-glucose required for P450 catalysis and NADPH regeneration in the whole-cell reaction of DDA hydroxylation may interfere with the detection results of GOase M3-5 [13]. If the Fe 2+/3+ /Cu 2+ ratio in the solution is too high, the Cu 2+ binding of GOase M3-5 would be hindered, leading to a loss of activity. Meanwhile, the activity of GOase M3-5 towards D-glucose may generate false positive results. Therefore, this HTS method needs further improvement before it can be applied for P450-directed evolution in the whole-cell reaction system.
Moreover, the activity of catalytic C-H activation is mediated by the active P450 formed after the binding with heme. In the catalytic process, heme plays the key role of electron transfer, that is, it receives electrons from FMN and attacks the closest C-H bond on the substrate [14][15][16]. In engineered systems with P450 overexpression, the heme synthesized by the natural metabolism is often insufficient to match the massive amount of P450. The exogenous addition of 5-aminolevulinic acid (5-ALA) as a precursor of heme has been shown to improve the catalytic efficiency of cells overexpressing P450 enzymes [17,18]. However, E. coli generally has a low utilization rate of exogenous 5-ALA, strengthening the heme supply by metabolic engineering, which may therefore be a viable strategy for improving P450 performance.
In our previously developed ω-AmDDA-producing E. coli strain [6], the chimeric P450 enzyme (cyp153a-ncp G307A ) constructed by fusing the monooxygenase CYP153a G307A mutant from Marinobacter aquaolei [9,19] and the reductase domain of P450 BM3 from Bacillus megaterium [20] was used. In order to further improve the ω-AmDDA bioproduction efficiency, in this study, the P450-catalyzed terminal hydroxylation was enhanced by both the directed evolution of cyp153a-ncp G307A and the engineering of the heme synthesis pathway. A modified ABTS colorimetry-based HTS method was established and used for the directed evolution of the chimeric P450 construct, selecting mutants with enhanced catalytic activity. Meanwhile, the heme supply in the engineered strain was enhanced to provide sufficient active P450 by properly strengthening the heme synthesis pathway. Finally, the efficiency of this strategy was examined in ω-AmDDA bioproduction.

Establishment of the High-Throughput Screening Method
The key to successful directed evolution lies in the availability of an efficient and reliable HTS method. The reported P450 enzyme activity assays are mostly based on NAD(P)H colorimetry (Figure 2a) [21], which is, however, unable to distinguish the regioselectivity of the enzyme. In ω-AmDDA biosynthesis, the highly selective terminal hydroxylation of DDA is a premise. Alternatively, p-nitrophenol colorimetry can be used to reflect the terminal hydroxylation activity of P450 using ω-p-nitrophenoxycarboxylate acids (pNCA) as artificial substrates (Figure 2b) [22]. However, the mutants selected with improved activity for the artificial substrates may not have the desirable performance for the target substrate. Recently, Weissenborn et al. [12] constructed an ABTS colorimetric method by coupling the galactose oxidase mutant GOase M3-5 capable of the specific oxidation of terminal fatty acid hydroxylates with horseradish peroxidase (HRP) (Figure 2c). In this HTS method, the site-specific catalytic activity of fatty acids can be measured and compared. Therefore, the GOase M3-5 -based ABTS colorimetry seems to be applicable for the screening of P450 mutants with enhanced DDA terminal hydroxylation activity.
To validate the feasibility of this HTS method for the directed evolution of P450 towards higher DDA terminal hydroxylation activity, the correlation between the absorbance and ω-OHDDA amount was first examined (Figure 3a). Subsequently, cyp153a-ncp (WT) and cyp153a-ncp G307A (M1), with a known activity difference, were tested as the low and high enzyme activity conditions, respectively, to further verify the feasibility of this method in the whole-cell catalysis system. Considering the Fe 2+/3+ and Cu 2+ dependence of cyp153a-ncp and GOase M3-5 , respectively, the effect of Fe 2+/3+ addition on the activity of GOase M3-5 and, thus, on the results of ABTS colorimetry was investigated (Figure 3c). The addition of 0.15 mM Fe 2+ at the protein induction stage was found to be the best, with good consistency with the HPLC results. In addition, the whole-cell reaction system contained 1% (w/v) D-glucose for the D-glucose dehydrogenase (GDH)-mediated NADPH regeneration to support the hydroxylation reaction, while GOase M3-5 , as a glucose oxidase, has D-glucose oxidation activity and could generate H 2 O 2 in this process [13], which may interfere with the colorimetric reaction and lead to false positive results. The exclusion of D-glucose from the reaction system led to limited ω-OHDDA formation, which was below the detection limit of this colorimetric method (Figure 3b). To avoid the interference of D-glucose, we tried to replace the GDH system with other NADPH regeneration systems, including ICD [23], GDHA [24], FDH [25], FDH-PNTAB [26,27] and FDH-STHA [26,27]. However, all those NADPH regeneration systems were not as effective as the GDH system (Supplementary Materials: Figure S1). tion of terminal fatty acid hydroxylates with horseradish pero this HTS method, the site-specific catalytic activity of fatty compared. Therefore, the GOaseM3-5-based ABTS colorimetry the screening of P450 mutants with enhanced DDA terminal Using ω-pnitrophenoxycarboxylate acids (pNCA) as the artificial substrate, the terminal hydroxylation activity of P450 is assayed by measuring the yellow p-nitrophenoxy ions generated upon the dissociation of the product. (c) Using a mutant of galactose oxidase (GOase M3-5 ) with the specific oxidation activity of terminal fatty acid hydroxylates, the terminal hydroxylation activity of P450 could be assayed by quantifying the actual product ω-OHDDA via measuring the H 2 O 2 generated during its oxidation to aldehyde by GOase M3-5 . The colorimetric measurement of H 2 O 2 is enabled by using horseradish peroxidase (HRP) and diammonium 2,2 -azino-bis(3-ethylbenzothiazoline-6-sulfonate) (ABTS). ABTS· + , the green-blue stable radical cationic chromophore formed by oxidation of ABTS.
Fortunately, GOase M3-5 was much more sensitive to ω-OHDDA than D-glucose ( Figure 3b). Using the absorbance of the buffer containing D-glucose as the background value, the result calculated by subtracting the background value from the absorbance value of the reaction product (I = max (I ω-OHDDA − I Buffer )) had a linear relationship with the concentration of ω-OHDDA in the range of 0.2-0.8 mM (Figure 3d). In this way, the false positive result caused by D-glucose in the system could be avoided. This modified HTS method was named ABTS 2.0 colorimetry. cules 2023, 28, x FOR PEER REVIEW Figure 3. Establishment and optimization of the high-throughput screening m ω-OHDDA standard reagents were determined by ABTS colorimetry. ΔOD420, t ence between the absorbance of the experimental group and the background at ΔOD420 = max (IE − IB). (b) Results of ABTS colorimetry for the reaction system cyp153a-ncp (WT). The reaction was performed with a 50 g cell wet weight (cw sodium phosphate buffer (100 mM, pH 8.0) with or without 1% (w/v) D-glucos addition at different stages (protein induction stage and whole-cell reaction sta of the ABTS colorimetry method, as shown by its consistency with the HPLC de C1-C6 stands for conditions 1-6 with different iron addition strategies. The y-ax activity of B1-1 (E-M1-3) containing M1 (cyp153a-ncp G307A ), as compared to B WT (cyp153a-ncp). +, 0.15 mM; ++, 0.5 mM. (d) Correlation between the maxim between the absorbance value of the product and the background and the prod absorbance value at 420 nm; max (), the maximum value of a set. The reactio sodium phosphate buffer (100 mM, pH 7.5), 25 μL whole-cell reaction product,

Directed Evolution of cyp153a-ncp G307A (M1)
The P450 mutant library was constructed with M1 as the parent by PCR (epPCR), and the mutation rate was controlled to 1-2 bp per kb. fragments were cloned into the expression plasmid E-M1-3 [6], and 15 m as the substrate for the HTS of the library. The mutants with increase duction were preliminarily screened by ABTS 2.0 colorimetry and confi order to ensure the accuracy of the screening, the reaction solution of t transformation was diluted to let the ω-OHDDA concentration fall in th mM before the measurement with ABTS 2.0 colorimetry. Among 1200

Directed Evolution of cyp153a-ncp G307A (M1)
The P450 mutant library was constructed with M1 as the parent by using error-prone PCR (epPCR), and the mutation rate was controlled to 1-2 bp per kb. The mutant gene fragments were cloned into the expression plasmid E-M1-3 [6], and 15 mM DDA was used as the substrate for the HTS of the library. The mutants with increased ω-OHDDA production were preliminarily screened by ABTS 2.0 colorimetry and confirmed by HPLC. In order to ensure the accuracy of the screening, the reaction solution of the whole-cell biotransformation was diluted to let the ω-OHDDA concentration fall in the range of 0.2-0.8 mM before the measurement with ABTS 2.0 colorimetry. Among 1200 clones, 5 mutants with improved activity were obtained ( Table 1). The mutant with R14R synonymous mutation showed a 35.9% higher activity than M1 (Figure 4a), and the yield of ω-OHDDA from 15 mM DDA reached 71.1% within 4 h. Because the R14 site is located at the Nterminus of P450 and the change in the codon preference near the N-terminus was reported to have a great impact on protein expression [28,29], the SDS-PAGE analysis of the mutants was conducted. The result showed an improvement in the soluble expression of the protein after the R14R synonymous mutation (Figure 4b,c), which was possibly due to the slight increase in the codon preference (from 0.36 to 0.37) and the moderate reduction in the translation rate (from 55,270 to 32,498), as calculated by the ribosome binding site (RBS) calculator [30,31] [https://salislab.net/software/ (accessed on 10 January 2023)].
a Mutants screened from the first round were named as "A + number", and mutants screened from the second round were named as "B + number". b These results were determined by the HPLC measurement of ω-OHDDA production, and all biotransformation reactions were performed in triplicate. The relative activities of the mutants obtained are given as the fold increase relative to ω-OHDDA production by M1.

Clones a Amino Acid Substitutions (Synonymous Muta
a Mutants screened from the first round were named as "A + number", and mu the second round were named as "B + number". b These results were determ measurement of ω-OHDDA production, and all biotransformation reactions triplicate. The relative activities of the mutants obtained are given as the fold in OHDDA production by M1. In the second round, cyp153a-ncp G307A/R14R (M2) was used as the pa with improved activity was screened out from 800 clones (Table 1). Th In the second round, cyp153a-ncp G307A/R14R (M2) was used as the parent, and 1 mutant with improved activity was screened out from 800 clones ( Table 1). The D629G mutation delivered 41.3% activity improvement (Figure 4d), and the yield of ω-OHDDA from 15 mM DDA reached 85.8% within 4 h. Because the D629 site was located in the FMN binding domain of the P450 reductase domain, we speculated that the enhancement of P450 activity in mutant M3 (cyp153a-ncp G307A/R14R/D629G ) might result from the improvement of electron transfer efficiency. The determination of the kinetic constant confirmed that the increase in catalytic activity was due to the increase in turnover numbers rather than the increase in substrate affinity ( Table 2). This result highlights the important role of electron transfer in P450-catalyzed reactions, which is in accordance with previous reports [8,10].

Molecular Simulation Analysis
In order to further explore the molecular mechanism of the enhanced catalytic activity in the D629G-containing mutant M3, molecular simulation analysis was carried out. Considering that the D629 site is located in the P450 reductase domain, the P450 reductase domain model (including the FAD, NAP and FMN binding domains) was first constructed through homologous modeling. In this process, we paid special attention to the arrangement of domains based on the direction of electron transmission: NADPH → FAD → FMN (Figure 5a). The FMN binding domain referred to the 1bvy model (PDB code: 1bvy [14]), with a sequence homology of 100%, the FAD and NAP binding domains referred to the 4dqk model (PDB code: 4dqk [32]), with a sequence homology of 100%, and the splicing of the three modules referred to the 1amo model (PDB code: 1amo [33]), with the same domains and similar structures (Figure 5a). The evaluation results using MolProbity [34] [http://molprobity. biochem.duke.edu/ (accessed on 28 November 2022)] and SAVES [35] [https://saves.mbi. ucla.edu/ (accessed on 28 November 2022)] demonstrated the high accuracy of this model. The evaluation results using MolProbity showed that 93.2% (549/589) of all residues were in favored (98%) regions, and 97.1% (572/589) of all residues were in allowed (>99.8%) regions ( Figure S2), the evaluation results using VERIFY3D showed that this model has 89.17% of the residues with an averaged 3D-1D score >= 0.2, and the evaluation results using the Ramachandran plot showed that this model has 89.7% (468/522) of the residues in most favored (>90%) regions ( Figure S3). The docking results of FAD, NADPH and FMN molecules with this model showed that the D629G mutation site is located in the FMN binding domain and near the cofactor binding pocket (Figure 5b). When the acidic aspartate was substituted by the neutral glycine, the surface electrostatic potential of the cofactor binding pocket was increased (Figure 5c), which was favorable for the negatively charged phosphate group of FAD in approaching and competing with the O1P of FMN for A627. The formation of a new hydrogen bond with a shorter distance (2.2 Å) to replace the original long hydrogen bond (3.5 Å) (Figure 5d,e) allowed for the rotation of FMN to a certain angle for the formation of new hydrogen bonds with N595 (2.9 Å) and S628 (3.0 Å), respectively (Figure 5f). Meanwhile, the distance between the rotated FMN and FAD was shortened from 9.1 Å to 8.3 Å (Figure 5g), and the shorter distance was conducive to the transfer of electrons from FAD to FMN [10,36,37], thereby improving the catalytic efficiency of P450. In addition, the performance of the mutant M3 was investigated for other fatty acids of different lengths, finding that the enzyme activities were improved for all of the substrates tested (Table 3), while retaining the specificity of terminal hydroxylation (Figures S4 and S5). This result showed that the mutant generated by the increasing electron transfer efficiency improved the catalytic performance in a substrate-independent manner, implying wide applications of such mutants. This also provides a reference for the modification of P450 and other enzymes with similar electron transfer domains.

Enhancement of the Heme Synthesis Pathway
Only when the heme domain of P450 binds to heme can the enzyme catalyze C-H activation [14][15][16]. Therefore, the adequate supply of heme in cells is essential to increasing the proportion of active P450. However, the heme synthesized by E. coli is limited (Figure 6a). In order to improve its synthesis, many optimization strategies have been reported. Weng et al. [38] overexpressed the hemA, hemB, hemC, hemD, hemE, hemF, hemG, hemH and hemL genes using the pUC19-hemAL, pACYCDuet-2-hemBCDE and pRSFDuet-2-hemFGH plasmids and increased the heme synthesis to 0.82 mg/L in E. coli. Zhao et al. [18] overexpressed the hemA and hemL genes on the pCDF-hemAL, hemB, hemC and hemD genes on the pRSF-hemBCD, hemE, hemF, hemG and hemH genes on pET-hemEFGH and knocked out the lactic acid-and acetate-forming genes (ldhA and pta) and the yfeX gene to enhance the supply of 5-ALA precursors and prevent heme degradation, respectively, which, together, increased the heme synthesis to 6.6 ± 0.2 mg/L. It has also been reported that optimizing the expression levels of the HemB, HemG and HemH enzymes alone promoted the transformation from 5-ALA to heme [39]. Ge et al. [17] found that the overexpression of the hemB gene alone led to a slight increase in the accumulation of 5-ALA, while the co-overexpression of the hemB, hemG and hemH genes had no significant impact on the production of 5-ALA. Significant heme accumulation was observed in both cases. Based on these reports, we tested three pathway engineering strategies and compared their efficiency: 1. Overexpressing hemB, hemC, hemD and hemE on a low-copy plasmid (pAC-hemBCDE), and overexpressing hemF, hemG and hemH on a high-copy plasmid (pRS-hemFGH) (B1-1-1); 2. Overexpressing hemB on a low-copy plasmid (pAC-hemB) (B1-1-2); 3. Overexpressing hemB, hemC and hemD on a high-copy plasmid (pRS-hemBCD), overexpressing hemE, hemF, hemG and hemH on a low-copy plasmid (pAC-hemEFGH) (B1-1-3) and deleting the yfeX gene (B1-1-3+∆yfeX).
It turned out that strategy 2, namely, overexpressing hemB on low-copy plasmid, led to the best result, which increased the ω-OHDDA yield of B1-1-2 by 7.0% (Figure 6b). Then, we compared the effect of integrative HemB expression driven by promoters of different strengths and found medium-intensity overexpression to be the best, which increased the ω-OHDDA yield of B2-2 by 20.7% (Figure 6b). In addition, it was found that the single deletion of yfeX had little effect on the product yield in B1-1, while its deletion in B1-1-3 with a background of strengthened heme synthesis improved the product yield. This result was suggested only when heme was massively synthesized, blocking its degradation-exerted positive effect. Therefore, we tentatively deleted yfeX from strain B1-1-2 with hemB upregulation and found a 24.6% increase in the yield of ω-OHDDA. These results indicated that the enhancement of the intracellular heme supply could effectively increase the proportion of active P450 ( Figure S6), thereby increasing the production of related products. It turned out that strategy 2, namely, overexpressing hemB on low-copy plasmid, led to the best result, which increased the ω-OHDDA yield of B1-1-2 by 7.0% (Figure 6b). Then, we compared the effect of integrative HemB expression driven by promoters of different strengths and found medium-intensity overexpression to be the best, which increased the ω-OHDDA yield of B2-2 by 20.7% (Figure 6b). In addition, it was found that the single deletion of yfeX had little effect on the product yield in B1-1, while its deletion in B1-1-3 with a background of strengthened heme synthesis improved the product yield. This result was suggested only when heme was massively synthesized, blocking its degradation-exerted positive effect. Therefore, we tentatively deleted yfeX from strain B1-1-2 with hemB upregulation and found a 24.6% increase in the yield of ω-OHDDA. These results indicated that the enhancement of the intracellular heme supply could effectively increase the proportion of active P450 ( Figure S6), thereby increasing the production of related products.

Expression, Purification and SDS-PAGE Analysis of Galactose Oxidase
The BL21+M3-5 strain was cultured overnight in 5 mL Luria-Bertani broth (LB, containing 50 mg/L Kanamycin) at 37 • C and 220 rpm as a seed culture, which was transferred to fresh LB (containing 50 mg/L Kana) with an inoculation volume of 5% (v/v) for further cultivation at 37 • C and 220 rpm until the OD 600 reached 0.6. For the protein induction, 0.1 mM IPTG and 0.5 mM CuSO 4 were added and cultured at 20 • C and 180 rpm for 10-16 h. The cells were collected by centrifugation and then washed and resuspended to a 50 g cell wet weight (cww)/L with sodium phosphate buffer (100 mM, pH 7.5) for ultrasonic cell disruption (200 W~400 W power, 90 cycles of 4 s fragmentations and 4 s intervals). After centrifugation, the GOase M3-5 crude enzyme was obtained as the supernatant, which was loaded onto 10% sodium dodecyl sulfate polyacrylamide gel for electrophoresis analysis and used for ABTS 2.0 colorimetry.

Cloning, Expression and Purification of P450
The construction of the P450 mutant library was carried out using epPCR. The plasmid E-M1-3 carrying the cyp153a-ncp G307A gene was used as the template to generate a random library through epPCR. Mutagenesis was performed on the full sequence of cyp153ancp G307A , using the CYPCPR-Gibson-F/CYPCPR-Gibson-R primers (Table S1). The epPCR reaction mixture contained 0.1~0.2 ng plasmid template, 0.2 mM dNTPs, 0.1 mM MnCl 2 and 5 U Taq DNA polymerase. The mutated cyp153a-ncp G307A was ligated to E-M1-3 [6] through the Gibson assembly method to replace cyp153a-ncp G307A on the original plasmid.
Expression of P450: The B1-1 [6] strain was used as the expression host. The preliminary screening was carried out in 96 deep-well plates. Single colonies were inoculated to 300 µL LB (containing 100 mg/L Amp) and incubated at 37 • C and 220 rpm for more than 10 h, and then 100 µL fresh seed solution was transferred to 24-well plates containing 2 mL Terrific broth (TB, containing 100 mg/L Amp) and cultured at 37 • C and 220 rpm until the OD 600 reached 0.8~1. For the protein induction, 0.1 mM IPTG, 1% (w/w) trace element stock solution (containing 0.15 M Fe 2+ ) [49], 0.5 mM 5-ALA and 65 mg/L VB1 were added. After culturing at 24.5 • C and 180 rpm for 10-16 h, the cells were collected for the whole-cell biotransformation and purification of P450.

Whole-Cell Biotransformation
The cells were washed with sodium phosphate buffer (100 mM, pH 7.5). The reaction was performed with 50 g cww/L of resting cells in sodium phosphate buffer (100 mM, pH 8.0) containing 1% (w/v) D-glucose, NH 3 ·H 2 O/NH 4 Cl (200 mM, NH 3 ·H 2 O:NH 4 Cl = 1:10) and 15.0 mM DDA (2% DMSO) for 2-8 h. The temperature was maintained at 30 • C, the agitation speed was maintained at 220 rpm, the pH was maintained at 7.5-8.0 and the concentration of D-glucose was maintained at 0.5-1% (w/v) throughout the biotransformation process. For rescreening, 1% (w/w) trace element stock solution (containing 0.5 M Fe 2+ ) [49] was added during the reaction. All biotransformation reactions were performed in triplicate.
For the GC/MS analysis, HCl was added to stop the whole-cell biotransformation reaction, and then 1 mM internal standard (DDA for the C 10:0 and C 13:0 substrates; tridecanoic acid for the C 14:0 and C 16:0 substrates) was added for product quantification. The reaction mixtures were extracted twice with 0.5 mL tert-butyl methyl ether. The organic phases were collected, dried with MgSO 4 (anhydrous) and evaporated. Samples were resuspended in 40 µL of 1% trimethylchlorosilane in N,O-bis(trimethylsilyl) trifluoroacetamide and incubated at 75 • C for 30 min for derivatization. The samples were analyzed on a Shimadzu GCMS QP2010SE instrument (Tokyo, Japan) equipped with a Shimadzu SH-I-5Sil MS column (30 m × 0.25 mm × 0.25 µm, Tokyo, Japan), with helium as the carrier gas (flow rate, 0.69 mL/min; linear velocity, 30 cm/s). Mass spectra were collected using electrospray ionization. The injector and detector temperatures were set at 250 • C and 285 • C, respectively. For the analysis of the C 10:0 -C 13:0 fatty acids, the column oven was set at 130 • C for 2 min, raised to 250 • C at a rate of 10 • C/min, held at isotherm for 3 min and then raised to 300 • C at 40 • C/min. For the C 14:0 and C 16:0 compounds, the temperature was maintained at 180 • C for 1 min, raised to 300 • C at 8 • C/min and held at isotherm for 5 min. Reaction products were identified by their characteristic mass fragmentation patterns [52].

Kinetic Analysis of P450
DDA was dissolved in DMSO and used at different concentrations (0.05-1 mM) as the substrate. The reaction was conducted with purified P450 in sodium phosphate buffer (100 mM, pH 7.5) containing 0.5 mM NADPH. The Epoch 2 microplate reader (BioTek Instruments, Inc, Winooski, VT, USA) was used to measure the absorbance value of NADPH at 340 nm and 30 • C every min [21]. The initial rate data were fitted nonlinearly with the Michaelis Menten equation to obtain the kinetic constant, and then the conversion number k cat and the catalytic efficiency k cat /K m were calculated.

Molecular Simulation Analysis of P450
Preparation of receptor proteins. According to the sequence and structure of the cyp153a-ncp reductase domain, protein models with a high sequence homology or similar structures in the PDB library [https://www.rcsb.org/ (accessed on 19 January 2022)] were selected as templates, and the Modeller 10.1 tool [53][54][55] was used to construct P450 reductase domain models through homologous modeling. The results are visualized in PyMOL. The MolProbity [34] [http://molprobity.biochem.duke.edu/ (accessed on 28 November 2022)] and SAVES [35] [https://saves.mbi.ucla.edu/ (accessed on 28 November 2022)] tools were used to evaluate the results, and the model with high scores was selected as the standard model of the receptor protein.
Molecular docking and analysis: The Grid tool of the AutoDockTools software was used to generate a Grid Box with an appropriate size containing all coenzymes binding pockets. The three coenzymes were successively docked by the AutoDock4 software. The configuration with the lowest binding energy and the coenzyme near the theoretical binding pocket was selected from the docking results and loaded into PyMOL software to analyze the chemical bonds between the coenzymes and the receptor protein.

Conclusions
In order to enhance the terminal hydroxylation activity of DDA catalyzed by P450, P450-directed evolution and heme synthesis enhancement were combined in this study to improve its soluble expression and enhance the electron transfer efficiency. The application of this strategy in the biosynthesis of ω-AmDDA generated an efficient whole-cell biocatalyst for PA12 monomer bioproduction. The replacement of the original P450 chimera M1 with the newly created mutants M2, with improved soluble expression, and M3, with improved expression and activity, enhanced the ω-AmDDA yields by 21.5% and 35.6%, respectively. The subsequent replacement of NAD + -dependent BsADH with FAD-dependent AlkJ and RBS engineering accelerated the conversion of ω-OHDDA to ω-AmDDA, which improved the ω-AmDDA yield by 136.7%. Strengthening the heme synthetic pathway led to a further 17.3% improvement in the ω-AmDDA yield. Finally, the whole-cell biocatalyst produced 9.36 mM ω-AmDDA (2.02 g/L), which was 227.0% higher than that of the original strain [6]. The HTS method established, the hot mutation sites identified in the P450 reductase domain and the strategy used to enhance the supply of heme in this work would provide helpful hints for engineering other P450-involved bioprocesses.